Tee-Ann Teo

The Geometrical Comparisons of RSM and RFM for FORMOSAT-2 Satellite Images

In this paper, we compare the geometrical performance between the rigorous sensor model (RSM) and rational function model (RFM) in the sensor modeling of FORMOSAT-2 satellite images. For the RSM, we provide a least squares collocation procedure to determine the precise orbits. As for the RFM, we analyze the model errors when a large amount of quasi-control points, which are derived from the satellite ephemeris and attitude data, are employed. The model errors with respect to the length of the image strip are also demonstrated. Experimental results show that the RFM is well behaved, indicating that its positioning errors is similar to that of the RSM.

Building reconstruction from LIDAR data and aerial imagery

This paper presents a scheme for building detection and reconstruction by merging LIDAR data and aerial imagery. In the building detection part, a region-based segmentation and object-based classification are integrated. In the building reconstruction, we analyze the coplanarity of the LIDAR point clouds to shape roofs. The accurate positions of the building walls are then determined by integrating the edges extracted from aerial imagery and the plane derived from LIDAR point clouds. The three dimensional building edges are thus used to reconstruct the building models. In the reconstruction, a patented method SMS (Split-Merge-Shape) is incorporated. Having the advantages of high reliability and flexibility, the SMS method provides stable solution even when those 3D building lines are broken. LIDAR data acquired by Leica ALS 40 and aerial images were used in the validation. Experimental results indicate that the successful rate for building detecition is higher that 81%. The positioning for buildings may reach sub-meter accuracy.

DEM-Aided Block Adjustment for Satellite Images With Weak Convergence Geometry

To acquire the largest possible coverage for environmental monitoring, it is important in most situations that the overlapping areas and the convergent angles of respective satellite images be small. The traditional bundle adjustment method used in aerial photogrammetry may not be the most suitable for direct orientation modeling in situations characterized by weak convergence geometry. We propose and compare three block adjustment methods for the processing of satellite images using the digital elevation model (DEM) as the elevation control. The first of these methods is a revised traditional bundle adjustment approach. The second is based on the direct georeferencing approach. The third is a rational function model with sensor-oriented rational polynomial coefficients. A collocation technique is integrated into all three methods to improve the positioning accuracy. Experimental results indicate that using the DEM as an elevation control can significantly improve the geometric accuracy as well as the geometric discrepancies between images. This is the case for all three methods. Moreover, the geometric performance of the three methods is similar. There is a significant improvement in geometric consistency between overlapping SPOT images with respect to single image adjustment for steep areas.

Bias Compensation in a Rigorous Sensor Model and Rational Function Model for High-Resolution Satellite Images

This paper presents three bias-compensated models for the geometric correction of high-resolution satellite images. The proposed models include the bias-compensated rigorous sensor model (RSM) in the orbital space, the bias-compensated RSM in the image space, and the bias-compensated rational function model (RFM) in the image space. The RSM and RFM use the on-board data and sensor-oriented rational polynomial coefficients (RPCs) provided in imagery metadata, respectively. Test images include QuickBird, WorldView-1, and WorldView-2 Basic images. Experimental results indicate that the bins-compensated RSM using the zero order polynomials function in the orbital space provides higher accuracy. A comparison of the bias-compensated RSM and RFM in the image space shows that these models behave similarly, and the maximum difference in root-mean-square error is less than 0.1 m. These results show that all the proposed methods obtain accuracy of better than 1 pixel, except for the translation in the image space.

Lidar-based change detection and change-type determination in urban areas

Change detection of objects, such as buildings, is essential for map updating. Traditionally, detection is usually performed through spectral analysis of multi-temporal images. This article proposes a method that employs multi-temporal interpolated lidar data. The objective of this study is to perform change detection and change-type determination via geometric analysis. A shape difference map is generated between the digital surface models in two different time periods. The areas with small shape differences are treated as non-changed areas and are excluded from the segmentation. The objects properties are then applied to determine the change types. Experimental results demonstrate that the proposed scheme achieves accuracy as high as 80%. Most of the errors from this study occurred in small or vegetation areas.

Reconstruction of building models with curvilinear boundaries from laser scanner and aerial imagery

This paper presents a scheme to detect building regions, followed by a reconstruction procedure. Airborne LIDAR data and aerial imagery are integrated in the proposed scheme. In light of the different buildings, we target the ones with straight and curvilinear boundaries. In the detection stage, a region-based segmentation and object-based classification are integrated. In the building reconstruction, we perform an edge detection to obtain the initial building lines from the rasterized LIDAR data. The accurate arcs and straight lines are then obtained in the image space. By employing the roof analysis, we determine the three dimensional building structure lines. Finally, the Split-Merge-Shape method is applied to generate the building models. Experimental results indicate that the success rate of the building detection reaches 91%. Among the successfully detected buildings, 90% of the buildings are fully or partially reconstructed. The planimetric accuracy of the building boundaries is better than 0.8m, while the shaping error of reconstructed roofs in height is 0.14 m.

BIM-oriented indoor network model for indoor and outdoor combined route planning

Emergency response and pedestrian route planning rely highly on indoor-outdoor geospatial data and a network model of the data; however, indoor geospatial data collection is time-consuming. Several studies have used the architecture, engineering, and construction (AEC) model to generate indoor network models. These models are subject to the input data types, and the attributes of interior building objects are usually incomplete; hence, the integration of building information modeling (BIM) and geographic information systems (GIS) can benefit indoor-outdoor integrated applications. To achieve data interoperability, an open BIM standard called Industry Foundation Classes (IFC) is maintained by buildingSMART. In this study, we propose a multi-purpose geometric network model (MGNM) based on BIM and explore the strategy of indoor and outdoor network connections. To achieve the goals, the IFC-to-MGNM conversion includes the following: (1) extraction of building information from IFC, (2) isolation of the MGNM information from the aforementioned building information, and (3) build up the topological relationships of MGNM into GIS Geodatabase. In addition, an entrance-to-street strategy is proposed to connect indoor networks, entrances, and outdoor networks for detailed route planning. The experimental results indicate that the MGNM could be generated from BIM automatically and applied to connect indoor and outdoor features for the multi-purpose application. Two use-case scenarios were developed to validate the proposed methods. Compared to actual distance, the relative error was improved by 5.1% and 65.5% in the horizontal and vertical routes, respectively, over the conventional indoor network model from 2D ground plan. In addition, the computational time taken by the proposed coarse-to-fine route planning method was 25% that of the traditional single-scale route planning method.

Line-based rational function model for high-resolution satellite imagery

The object-to-image transformation of high-resolution satellite images often involves a rational functional model (RFM). Traditionally, RFM uses point features to obtain the transformation coefficients. Since control lines offer greater flexibility than control points, this study proposes a new RFM approach based on linear features. The proposed methods include direct RFM and bias-compensated RFM using control lines. The former obtains the rational polynomial coefficients (RPCs) directly from control lines, whereas the latter uses sensor-orientated RPCs and control lines to determine compensated coefficients. The line-based RFMs include vector and parametric line representations. The experiments in this study analysed the effects of line number, orientation, and length using simulation and real data. The real data combined three-dimensional building models and high-resolution satellite images, such as IKONOS and QuickBird images. Experimental results show that the proposed algorithms can achieve pixel-level accuracy.

Pole-Like Road Object Detection From Mobile Lidar System Using a Coarse-to-Fine Approach

The requirements of three-dimensional (3-D) road objects have increased for various applications, such as geographic information systems and intelligent transportation systems. The use of mobile lidar systems (MLSs) running along road corridors is an effective way to collect accurate road inventories, but MLS feature extraction is challenged by the blind scanning characteristics of lidar systems and the huge amount of data involved; therefore, an automatic process for MLS data is required to improve efficiency of feature extraction. This study developed a coarse-to-fine approach for the extraction of pole-like road objects from MLS data. The major work consists of data preprocessing, coarse-to-fine segmentation, and detection. In data preprocessing, points from different trajectories were reorganized into road parts, and building facades alongside road corridors were removed to reduce their influence. Then, a coarse-to-fine computational framework for the detection of pole-like objects that segments point clouds was proposed. The results show that the pole-like object detection rate for the proposed method was about 90%, and the proposed coarse-to-fine framework was more efficient than the single-scale framework. These results indicate that the proposed method can be used to effectively extract pole-like road objects from MLS data.

Surface-Based Registration of Airborne and Terrestrial Mobile LiDAR Point Clouds

Light Detection and Ranging (LiDAR) is an active sensor that can effectively acquire a large number of three-dimensional (3-D) points. LiDAR systems can be equipped on different platforms for different applications, but to integrate the data, point cloud registration is needed to improve geometric consistency. The registration of airborne and terrestrial mobile LiDAR is a challenging task because the point densities and scanning directions differ. We proposed a scheme for the registration of airborne and terrestrial mobile LiDAR using the least squares 3-D surface registration technique to minimize the surfaces between two datasets. To analyze the effect of point density in registration, the simulation data simulated different conditions and estimated the theoretical errors. The test data were the point clouds of the airborne LiDAR system (ALS) and the mobile LiDAR system (MLS), which were acquired by Optech ALTM 3070 and Lynx, respectively. The resulting simulation analysis indicated that the accuracy of registration improved as the density increased. For the test dataset, the registration error of mobile LiDAR between different trajectories improved from 40 cm to 4 cm, and the registration error between ALS and MLS improved from 84 cm to 4 cm. These results indicate that the proposed methods can obtain 5 cm accuracy between ALS and MLS.